Towards Autonomy in Daily Living: A Formalism for Intelligent Assistive Feeding Systems Applicant PI, Tapomayukh Bhattacharjee Overview We propose to develop a design space framework and co-design methodology for the development of assistive self-care robot technologies that are informed by the social model of disability. Our model of assistive robots in the domain of self-care considers an individual's social and environmental context, coping processes and other factors that can affect independent functioning. Our design methods utilize embedded sensing to intelligently respond to these con- siderations. We speci?cally focus on assistive feeding tasks, proposing a formalism that enables a robotic system to feed a person with upper-extremity disability. Our guiding principle is that human-level interaction is feasible only if the robot itself relies on human-level semantics. We im- plement this principle by relying on data to learn and develop object-dependent control policies and timing models for acquiring and transferring a bite to a user at a proper time. The system's ob- server detects world states and arbitrator invokes different control policies based on these states. The tangible result will be an intelligent assistive feeding robot whose performance can generalize to different activities, adapt to user preferences, and recover from failures. Objectives and Relevance to NIH A design framework for assistive robots would provide for- malisms that let us address the fundamental challenge of designing robots that are responsive to context of use and support assisted self-care in a variety of social settings. We combine method- ologies from human-robot interaction, cognitive science, machine learning, robotics and haptics with user studies and our formalism to address the following research questions: (Q1) Mechanics of Feeding-Control Policies: How can control policies be designed for dexterous non-prehensile manipu- lation of deformable objects such as food? (Q2) Social Aspects of Feeding-Bite Timing: How should an assistive feeding robot decide the right timing for feeding a user? (Q3) Human-in-the-Loop: How can human-directed feedback be added into the loop for an autonomous assistive feeding system? The proposed work will allow users with upper-arm disabilities to use this system for intelli- gent assistance with daily feeding tasks. This can in turn help them increase their independence and autonomy making eating easier and more enjoyable. While we presently focus on this spe- ci?c application, the tools and insights we gain can generalize to the ?elds of robotic assistance and human-robot interaction across other activities of daily living and instrumental activities of daily living. Thus, our work is clearly motivated by the intent to improve the quality of health and life of the aging population and is very relevant to the theme of NIH. 1